Update to CPA 0.11.0 to add downscale stabilization

Question

Closed this issue 4 years ago · 0 comments

Update to CPA v0.11.0, requires updating CPA-HPA.

[v0.11.0] - 2020-02-28

Series of hooks for injecting user logic throughout the execution process.
- preMetric - Runs before metric gathering, given metric gathering input.
- postMetric - Runs after metric gathering, given metric gathering input and result.
- preEvaluate - Runs before evaluation, given evaluation input.
- postEvaluate - Runs after evaluation, given evaluation input and result.
- preScale - Runs before scaling decision, given min and max replicas, current replicas, target replicas, and resource being scaled.
- postScale - Runs before scaling decision, given min and max replicas, current replicas, target replicas, and resource being scaled.
New downscaleStabilization option, based on the Horizontal Pod Autoscaler downscale stabilization, operates by taking the maximum target replica count over the stabilization window.

Metrics from API now returns the entire resource definition as JSON rather than just the resource name.
Changed JSON generated to be in camelCase rather than snake_case for consistency with the Kubernetes API.
- Evaluation now uses targetReplicas over target_replicas.
- ResourceMetric now uses runType over run_type.
- Scale hook now provided with minReplicas, maxReplicas, currentReplicas and targetReplicas rather than their snakecase equivalents.
Metric gathering and hooks have access to dryRun field, allowing them to determine if they are called as part of a dry run.
Standardised input to metric gatherer, evaluator and scaler to take specs rather than lists of parameters, allowing easier serialisation for hooks.
Endpoint /api/v1/metrics now accepts the optional dry_run parameter for marking metric gathering as in dry run mode.
ResourceMetrics replaced with a list of Metric and a Resource.
/api/v1/metrics now simply returns a list of Metrics rather than a ResourceMetrics.